Cerenkov radiation from bunched electron beams

■ id d ac y

RESEARCH REPORTS DIVISION
NAVAL POSTGRADUATE SCHOOL
MONTEREY, CALIFORNIA 93943

bort Number NPS-61-83-003

IIBRARY

f e IS DIVISION

naval; : ate school

MONTEREY, CALIFORNIA 9394Q

NAVAL POSTGRADUATE SCHOOL

Monterey, California

CERENKOV RADIATION FROM

BUNCHED ELECTRON

BEAMS

F.

R.

Buskirk and J.R.

Neighbours

October 19 82

Revised April

1983

Technical

Report

Approved for public releaser distribution unlimited

repared for:
FEDDOCS hief of Naval Research

D 208.14/2:NPS-61 -83-003 rlington, Virginia 22217

NAVAL POSTGRADUATE SCHOOL
Monterey, California

Rear Admiral John Ekelund David Schrady

Superintendent Provost

The work reported herein was supported by the Chief
of Naval Research.

Reproduction of all or part of this report is authorized.
This report was prepared by:

UNCLASSIFIED

'ittUHlTY CL ASSlFlt ATIONOF Inlt> PtOE iWUm Dmlm tnlmtmj)

RE PORT DOCUMENTATION PAGE

1 REPONT NUMBER

NPS-61-83-003

GOVT ACCESSION NO

* TITLE (mnd Subitum)

Cerenkov Radiation from Bunched
Electron Beams

7. AUTMO«C«J

F. R. Buskirk and J. R. Neighbours

» PERFORMING ORGANIZATION NAME AND AOORESS

Naval Postgraduate School
Monterey, CA 93940

II. CONTROLLING OFFICE NAMC AND AOORESS

Chief of Naval Research
Arlington, Virginia 22217

rr

KEAD INS7 WUCTIONS
UtHiKh i UMI'LETING KOkM

J RECIPIENT'S CATALOG NUMBER

5 type of report * perioo coveaeo *,
Technical Report

• PERFORMING ORG. REPORT NUMftA

. _J

I CONTRACT OR GRANT NUMBERS

10 PROGRAM ELEMENT. PROJECT, TASK
AREA * iOHK UNIT NUMtfNS

62763N;RF68-342-800
N0001482WR20195

12. REPORT DATS

November 19 8 2

11 NUMBER OF PAGES

MONlTOAlNO ACCNCV NAMC A AOORESSflf dlttmtmnt from Controlling Olllcm)

<» SECURITY CLASS, (ol inn import)

Unclassified

It* DECLASSIFICATION' OOWNGRACTfiG-
SCHEDULE

It. DISTRIBUTION STATEMENT (ol «ii» Report;

Approved for public release; distribution unlimited

17. DISTRIBUTION STATEMENT (ol thm •».rr»cl antmtmd In Slock 20. II dlllmrmnt Itom Rmport)

1*. SUPPLEMENTARY NOTES

It- KEY WORDS (Contlnum on tmvmtmm mldm II n«C(li«ry «j Idmnllty oy block numbmr)

Cerenkov Radiation
Microwave Radiation
Bunched Electrons

20. ABSTRACT (Cominum on tmvmtmm mldm It nmemmsmry mnd Idmnllty 6r mloe* m~a>o«<i

Cerenkov radiation is calculated for electron beams which exceed
the velocity of radiation in a non dispersive dielectric medium.
The electron beam is assumed to be bunched as emitted from a
travelling wave accelerator, and the emission region is assumed to
be finite. Predictions include (a) emission at harmonics of the
bunch rate, (b) coherence of radiation at low frequencies (c) smear-
ing of the emission angle for finite emission regions, (d) explicit
evaluation of cower sroectrum in terms of bunch dimensions. The

do ,;

FORM
AN 7)

1473

EDITION OF I NOV «t IS OBSOLETE
S/N 0 102-0 14- 660 1

UNCLASSIFIED

SECURITY CLASSIFICATION OF THIS PAGE (Whmn Dmlm Sntmtmdi

UNCLASSIFIED

CfaCURITV CLASSIFICATION OF THIS PAGECWi^n Dmtm Entmrmd)

results apply to microwave emission from fast electrons in
air or other dielectrics.

DD Form 1473 (BACK)
S/N1 "SfolV-SeO! UNCLASSIFIED

SECURITY CLASSIFICATION OF THIS PAGE(TWi«n Dmtm Entm

TABLE OF CONTENTS

Page

Introduction 1

Calculation of Poynting Vector 3

Fourier Coefficients of the Current 6

Vector Potential 8

Radiated Fower 11

Cerenkov Angle 13

Discussion of Results 15

a. Effect of Pulse Size 15

b. Shearing of the Cerenkov Angle 16

c. Behavior at high Frequencies Related 16
to Pulse Parameters.

Concluding Remarks; 18

References 20

APPENDICES

A. Derivation of Cerenkov Radiation Al
for a Single Pulse of Charge

B. Derivation of Equation 7 Bl

C. Temporal Structure of the Electron CI
Pulse from a Travelling Wave

Linear Accelerator

D. Form Factors Dl

Distribution List

INTRODUCTION

The radiation produced by gamma rays incident on ordinary
dielectric materials such as glass was first discovered by
Cerenkov1 in 1934 and was described in terms of a charged
particle (electron) moving faster than light in the medium by
Frank and Tamm2 in 1937. A summary of work to 1958 is contained
in the treatise by Jelly-*. An important application is the
Cerenkov particle detector which is familiar in any particle
physics laboratory, and an early and crucial application occurred
in the discovery of the antiproton4- .

Because the distribution of intensity of Cerenkov radiation
is proportional to the frequency, the radiation at microwave
frequencies would be low unless beams are intense and bunched so
that coherent radiation by many electrons contributes . Danos^
in 1955 calculated radiation produced by a planar beam passing
above a dielectric interface and a hollow cylindrical beam passing
through a hole in a dielectric. Experimental and theoretical
investigations at microwave frequencies were reviewed by
Lashinsky^ in 1961.

This investigation was motivated by a recent renewed interest
which has included the study of stimulated Cerenkov radiation, in
which the electron may be in a medium consisting of a gas' or a
hollow dielectric resonator^* 9. Recent developments of
electron accelerators for applications such as free electron
lasers (FEL) have aimed toward high peak currents in bunches in
contrast to nuclear and particle physics applications, where low
peak but high average currents are desirable to avoid saturating

1

detectors. The high peak currents in the new accelerators should
yield enhanced Cerenkov radiation, as is calculated in this
paper .

CALCULATION OF THE POYNTING VECTOR

In the following derivation, we consider the Cerenkov

radiation produced in a dispersionless medium such as cases or

other dielectrics, by a series of pulses of electrons such as are

produced by a traveling wave electron accelerator (Linac). The

pulses or bunches are periodic, the total emission region is

finite and the bunches have a finite size.

In determining the radiated power, the procedure is to

calculate the Poynting vector from fields which are in turn

obtained from solutions of the wave equations for the potentials.

Since the current and charge densities entering into the wave

equations are expressed in fourier form the resulting fields and

radiated power also have fourier components. In the derivation, r

is the coordinate at. which the fields will be calculated, r i.s

the coordinate of an element of the charge which produces the

fields and n is a unit vector in the direction of r. We assume

that E(r,t) and B(r,t,) have been expanded in a fourier series;,

appropriate for the case where the source current 1.3 periodic.

Then we have

00

E (tt) -£ e"iaJt1 tf,a»)

oj=-°° (1)

and a corresponding expansion for B, where w is a discrete

_& — ^
variable and E and B are fourier series coefficients. The

poynting vector S is given by

S = -~E x~B

(2)

and it is easy to show that the average of S in a direction given

A

by a normal vector n i:

T oo

i r n.s at - i y. -lu.

-J /i\=— oo

oj ) x B(r,-(D)

0)=-»

(3)

where T is an integer multiple of the period of the periodic
current .

Letting c = (ye) "1/2 be the velocity of light, in the
medium, the wave equations for A,4> and their solutions are,

2

3 -*■,-*> -*■ »
V — 9 — 0\ A(r,t) = yj(r,t)

cz3tzi

I c2dt2i

'?ir,t) = 1/e p(r,t)

m

3 ,

A(r,t) =y D(r-r', t-t ') J(f',t ') dr 'dt '

«f> (r,t) = 1

e

D(r-?',t-t,)p(r' ,t')d3r'dt'

(4)

(5)

where the Green's function D is given by

D(r,t) = 1_ 5 (t-r/c)

4irr

(6)

The vector potential A(r, t) also can be developed in a fourier

series expansion of a form similar to (1) with an expression for

the fourier series coefficients given by
T
A(r,a>) = 1 r dtA(?,t)e1'

1 f dtA(r,t)

T I

■ill

d3r> j(r' rai)_l 1__ eiaJlr-r' ' /c

r-r'

(7)

Now if we assume that the observer is far from the source so
that jr j>> |r' | for regions where the integrand in (7) is important
we can let |r-r | = *r - n • r in the exponential and jr - r' = :
in the |r - r1 |~1 factor in (7), obtaining (where n = r/r)

A(r,o>) = Ji_ e^r/c r[fa3r'j;(r' ,a>)e-iiai/crn-t'

47rr JJJ (8)

The fourier series coefficients of the fields are obtained

from those for the vector potential (8) through the usual

relations B = V x A and E = -V<|> - 9A . Under the conditions

3t
leading to (3) the field fourier coefficients are-^:

-* - ^ -l ^

B(r,uj) = ico n x A ( r , oj )

c (9)

-* -i -A -* -k

E(r,co) = -c n x B(r,oj)

(10)

The poynting vector can now be found by using (9) and (10)
in expansions like (1) and then substituting in (2). However it
is more convenient to deal with the frequency components of the
radiated power by substituting (9) and (10) into the expression of
the average radiated power (3).

T 00

1 [ n-Sdt = 1 )
T J y ^—

oj n x A ( r , oo )

c (11)

FOURIER COMPONENTS OF THE CURRENT
The expression (7) for the fourier components of the vector
potential contains the fourier components of the current density.
Consequently it is necessary to examine the form of the current
and its fourier development. Assume the current is in the z
direction and periodic. If the electrons move with velocity v,
and we ignore for the moment the x and y variables, the charge or
current functions should have the general form

f(z,t) = £ elkzZ £ e ia)t f(kz,a») (12)

(13)

Z CO

Under the condition of rigid motion,
f (z,t) = f0 (z-vt)

it is easy to show that

,(k ,cj) = 6 . f (k ) , %

z co,kv~0z (14)

where

fvkj = e 1Kzz f0(z)dz

*v o Z —

z

(15)

Thus the restrictions of equation (13) reduce the two
dimensional fourier series of eq. (12) to essentially a one
dimensional series (14).

With (14) in mind, the current density associated with the
electron beam from a linear accelerator should be periodic in both
z, t, with a fourier series expansion, but the x and y dependence
should be represented by a fourier integral form:

4«o

J_(r,t) ~ vp(r,t) =

(2 TT)

— CO

dk dk
x y

f ,, \~ i (k-r - cat)
J dky ^_ e Po

k =-c
z

(k)
16)

where the fourier components of the charge density are

P,o (k) = dx dy - | dze

U7)

p0 (r) is p(r,t) evaluated at t = o and J is assumed to be in
the z direction. Note in eq. (16) that kz and ^ are both
discrete and from (14), oj = kzv.

VECTOR POTENTIAL

The results of the previous section can be applied to the
evaluation of the vector potential and in turn to the fields.

Let the infinite periodic pulse train be made finite,

extending from z = -Z ' to z = +Z ' and let 9 be the angle between n

and A. Then the cross product in (11) can be written

I A *t #"*" \ I a V ioor/c

n x A(r,oj) = smfl -7 — e

CO GO 7 '

f dx' / dy' /" dz' e"ifi^,a)/c

Jm J>™ =^>7 I

JSOO .= 00 — 2

k =-c

t-At2 dk / dk ) v0o (£)<$, e1K'r

(2n)z I x Jot *y k^=_^ - kzv/W (18)

But

Z* j. *

dx' J dy' / dz' eir'-(k ™/c)

-Z'

(2tt)2 5(kx - nxaj/c) 5 (k - nu/c)I(Z') ( 19 )

where

Z'

T,„n / , , i(k - n co/c)z' •>

I(Z') = / dz' e z z 2 sinGZ. (20)

J-Z' G

and G = kz - n2 go /c = 00 /v - nz co /c
And thus the cross product term is

/\ •*■ ^ 1 Tj icor/c

n x A(r,a>) I = sinG-^r e v£o(nxVc, n oi/c, o)/v)I(Z')

(21)

Note that. o> is a discrete variable but from 19, the
continuous variables kx and ky become evaluated at discrete
points .

Returning to (17), a more symmetric form may be obtained by
assuming rhat p0(r)# which is periodic in z with period Z, is
actually zero between the pulses. Denoting by p0 ' ( r) the charge
density of a single pulse, which is zero for z < o and z > Z the
integral on z can be written

z z

, -ik z ,-». f , -ik z ',■»> -, -ik z ' ,a
dz e z p0(r; = dz e z p0(r) = dz e z p0 (r)

J (22)

Then (17) the fourier coefficient-, of the charge density, becomes

oo

p0(k) =| |fjrd3r e"ik'r:0'(r) = |pl ik)

(23)

where pn'(k) is the three dimensional fourier transform of the
single pulse decribed by pQ' (r) .

Substituting these expressions into (21) gives a final simple
result, for the cross product, form:

|n x A(r,<o)l = sine H elajr/C(v/Z) Pe' (k)I(z')

47tr - (24)

where

I(Z' ) = | sin GZ'
G

G = oj/v - n co/c (25 )

k = (n oo/cn a)/c,oo/v)

The components of the Cerenkov E and B fields may now be found by
substituting (24) in (9) and (10).

10

RADIATED POWER
The frequency components of the average radiated power are
obtained by substituting (24) into (II). The negative: frequency
terms equal the corresponding positive frequency terms , yielding a
factor of 2 when the summation range is changed. Multiplying by
r^ converts to average power per unit solid angle, dP/dft ,
yielding

T 00

f§=r2i I" n-Sdt = r2l£V |nx A(?,M)| 2

00

=1

C

W(<o,n)

(26)

where w(a)/n) is defined to be

2
w (a) ,n) = — 2 g- sin 9 (v /Z ) | p 0' (kj j I (Z )
(4it)

(27)

W(o),n) is the power per unit solid angle radiated at the frequency
0), which is a harmonic of the basic angular frequency uj0 of the
periodic pulse train.

To find P , the total power radiated at. the frequency co ,
W is multiplied by dfi and integrated over solid angle. Note that
nz = cos 9, and as 9 varies, G changes according to (25),

11

w

ith dG = - (co/c) dn so that

d^ = dc{> (c/oj) dG

(28)

Noting that the integral over <£ yields 2"^ , we find the result for
the total radiated power at the frequency oj for all angles

G"

p = r_ _

2 2

}J_ 60 v^

4tt c

2 J

sin26 jPo(k) |2 I2(Z') 5 dG

(29)

G'

12

CERENKQV ANGLE
The remaining integral over G may now be examined. The
sin^ S and p factors may often be slowly varying compared to
the l2(z') factor, the latter being shown in Fig. 1. For large
Z', the peak in l2(z') becomes narrow, and if the integrand may
be neglected outside the physical range G'<G<G",

G

I,

I (Z ')dG = |4(Z ')

/

i, 2 sinGZ

'}

GZ'

'dG = 4tt

(30)

Then, evaluating the sin 9 factor and pQ(k) at the
point corresponding to G = 0, (which is cos 6 = nz = c/v) shows
that. 9 at. the peak of I(Z' ) is the usual Cerenkov angle 9C. We
thus obtain for large Z'

P = £- covsin 6 |p0(k)| 4ttZ'/Z

oj 4lT Cl ~° '

(31)

Now let 2Z'/Z = ratio of the interaction length to pulse spacing
N, the number of pulses. Also Z = v2 tt/ o^ or 2tt/z = '^0/v so
that, (in the large Z' limit),

2 ** 2

Poj = 4tt" 000°ovsin 9cIP=^k)i N.

(32)

To compare with usual formulations, (32) is divided by Nv to
obtain the energy loss per unit path length per pulse:

dE = y_
dx 4tt

o)co0 sin 8 I p0 (k) ,

, 1 2

(33)

13

If the pulse is in fact, a point charge, the fourier transform
Dq (k) reduces to q, the total charge per pulse and (33) is
very similar to the usual Cerenkov energy loss formula, where for
a single charge q. the radiation is continuous and the factor

oo cj in (33) is replaced by u) d go. In the present case the pulse

o

train is periodic at angular frequency ojq and the radiation is
emitted at the harmonic frequenies denoted by oj .

14

DISCUSSION OF RESULTS
Equation (29) and the approximate evaluation expressed as
(32) form the main results. Some consequences will now be noted.

a. EFFECT OF PULSE SIZE. The spatial distribution of the
charge in the pulse appears in p^lk) , which is the fourier
transform of the charge distribution. The peak of I2(2') in
figure 1 occurs at G = 0 or nz = c/v. Thus at the peak, co/v = nz oj/c
i\o that k, the argument of p' (k), is evaluated at (from 25)

s+S 0
"*■ A

k = noo/c

(34)

We may also define a charge form factor F(k)

o;<k) = qp(ki (3S)

The form factor F(k) is identically one for a point charge,
and for a finite distribution F(k) = 1 for k =o .

Furthermore F(k) must fall off as a function of k near the
origin if all the charge has the same sign. If the pulse were
spherically symmetric, F(k) would behave as elastic electron
scattering form factors defined for nuclear charge
distributions 1-1 . In that case, the mean square radius <r2>
of the charge distribution is given by the behavior of F(k) near
the origin.

F(k) ■*■ 1 - <r2> k2/6 (spherical pulse) (36)

15

b. SMEARING OF THE CERENKOV ANGLE. For a finite region over
which emission is allowed, namely if 2Z ' is finite, the function
I2(Z'), appearing in the integral in (29), will have a finite
width. Since the peak height is 4Z ' 2 and the area is 4tts ' ,
(30), we can assign an effective width 2T = area/ height - ^ I z '

or

r=7T/2z' (37)

Thus the radiation is emitted mainly near G = o (which
corresponds to 9= 9C) hut in a range Ag = +T . But from (25),

00 , 00 . . .

AC = — in7 = — a(cos9) so that there is a range in cos0 over
c ^ c

wr. ich emission occurs:

a / Q^ C 7T (38)

A (cos9) = - ^—r

00 ZZj

Note that the finite angular width of the Cerenkov cone angle
in (38) has the factor l/co , indicating that the higher harmonics
are emitted in a sharper cone.

c. 3EHAVI0R AT HIGH FREQUENCIES RELATED TO PULSE PARAMETERS.
To be specific let the charge distribution for a single pulse be
given by gaussian functions

1 .-* 2 2 ? 2 0 9

p0(r) = A exp(-x/a - y /a - z7b )

Then F(k) may be found

39)

F(k) = exp(-k 2a2/4 -k 2a2/4 -k 2b2/4)

y Z (40)

Beam pulse parameters could then be determined by measuring the
Cerenkov radiation. For example, fast electrons from an
accelerator in air will e ait with a 9C of several degrees
in which case kx and ky in (40) can be neglected, giving

F(k) = exp(-kz2b2/4) = exp[-co2b2/(4v2) ] (41)

The expected behavio:: of P as a function of go is shown

00

qualitatively in Fig. 2 as a linear rise at. low frequencies
followed by a fall off at. higher frequencies, the peak occurring
at

(o = v/b
m

(42)

Furthermore, a different behavior would be expected at. very
high frequencies. The formulation from the beginning represents
coherent, radiation from all charges, not only from one pulse, but.
coherence from pulse to pulse. F(k) then describes interference
of radiation emitted from different, parts of the pulse, but note
that expressions (29} and (32) will still be proportional to
q2 - n2e2 where n ±s the number of electrons in a pulse.
Thus the n^ dependence of P indicates coherence. But. above

CO

some high frequency go j_ such that oj . /c - 2\\ I % , where I is the

mean spacing of electrons in the cloud, the radition should switch

over to incoherent radiation from each charge and P should be

00

proportional to n. The incoherent radiation should then rise
again as a function of au .

17

CONCLUDING REMARKS

The general results presented here describe the Cerenkov
radiation produced by fast electrons produced by a linear
accelerator. For an S band Linac operating at about 3Ghz (10 cm
radiation) , the electron bunches are separated by 10 cm and would be
about 1 cm long at 1% energy resolution. Microwave Cerenkov
radiation is expected and has been seen in measurements at the Naval
Postgraduate School Linac.

Two types of measurements were made. In measurements of Series
A, an X-band antenna mounted near the beam path, oriented to
intercept the Cerenkov cone, was connected to a spectrum analyser.
Harmonics 3 through 7 of the 2.85 GHz bunch frequency were seen but
power levels could not be measured quantitatively. Harmonics 1 and 2
were below the wave guide cut off. In the series B measurements, the
electron beam emerged from the end window of the accelerator, and
passed through a flat metal sheet 90 cm downstream oriented at an
angle <j> from the normal to the beam. The metal sheet defined a
finite length of gas radiator, and reflected the Cerenkov cone of
radiation toward the accelerator but rotated by an angle 2<t> from the
beam line. A microwave X-band antenna and crystal detector with
response from 7 to above 12 GHz could be moved across the (reflected)
Cerenkov cone as a probe.

As mentioned earlier, the series A measurements showed the
radiation is produced at the bunch repetition rate and its harmonics.
Series B measurements performed with several antennas always
indicated a broadened Cerenkov cone with strong radiation occuring at
angles up to 10°, well beyond the predicted Cerenkov angle of 1.3°.

18

Since a broad band detector was used it was impossible to verify
the prediction (see eq . 38) that the broadening cf the cone should
depend on the harmonic number. However, it should be noted that
incoherent radiation by a beam of lu A at 9 c = 1.3° for a 1
meter path in air would be about 10~^ watts at microwave
frequencies so that observation of any signal by either method A
or B clearly demonstrated coherent radiation by the electron
bunches .

Many of the concepts were clearly notea by Jelly in his
treatise (Jelly^, Section 3.4 especially). The form factor was
noted but a general expression was not given. In fact, the form
factor quoted by Jelly represents the special case of a uniform
line charge of length L' with a projected length L=L'cos9c in
the direction of the radiation. The coherence of the radiation
from the bunch was noted but no broadening of the: cone nor
harmonic structure were developed.

Casey, Yen and Kaprielian-^ considered an apparently
related problem in Cerenkov radiation, in whj ch a single electron
passes through a dielectric medium, where a spatially periodic
term is added to the dielectric constant. The result is radiation
occurring even for electrons which do not exceed the velocity of
light in the medium, and at angles other than the Cerenkov cone
angle. The non-Cerenkov part of the radiation is attributed to
transition radiation.

In the present paper, the transition radiation associated
with the gas cell boundaries is included, and radiation appears
outside the Cerenkov cone.

19

If the electron velocity were lower so that v/c were close to but
less than unity, the peak in I would be pushed to the left in Fig.
1, such that cos 9 c = v/c would be larger that 1. But. the tails of
the diffraction function I would extend into the physical range 1 _<
cos 8 <_ - 1 , and this would be called transition radiation and he
ascribed to the passage of the electrons through the boundaries of
the gas cell. Now return to the case v/c > i, with the situation as
shown in Fig. 1. The radiation is then a combination of Cerenkov and
transition radiation. The formalism of Reference 12 does admit a
decomposition into the two types of radiation, but is inherently much
more cumbersome.

As a final remark, one might extend the analysis further in the
region near w ^ . Consider electron bunches emitted frcm a
travelling wave Linac, which could be 1 cm long spaced 10 cm apart.
Let these bunches enter the wiggler magnet of a free electron laser
(FEL) . Then, if gain occurs, the 1 cm bunches would be subdivided
into bunches of a finer scale, with the spatial r-cale appropriate to
the output wavelength of the FEL.13 If the (partially) bunched
beam from the FEL were passed into a gas Cerenkov cell, then the
observed radiation should be reinforced because cf partial coherence,
at the FEL bunch frequency and harmonics. This would lead to bumps
in the spectrum in the region near oj ^ .

20

ACKNOWLEDGEMENTS

This work was made possible by the Office of Naval Research
W. B. Zeleny provided much insight into the theory for point
charges, and the experiments performed by Ahmet Saglern, Lt . ,
Turkish Navy, provide preliminary substantiation of the work.

21

REFERENCES

1. Cerenkov, P., Dcklady Akad-Nauk, S.S.S.R., 2 451, (1934).

2. Frank, I.M. and Tamm, I., Doklady Akad-Nauk 3.S.S.R., L4 109,
(1937) .

3. Jelly, J. V., "Cerenkov Radiation and its Applications".
Pergammon Press, London, (1958).

4. Chamberlain, 0., E. Segre, C. Wiegand and T. Ypsilantis, Phys .
Rev. , 100 947, (1955) .

5. E'anos, M.J., J. Appl. Phys, 26_ 1 , (19 55).

6. Laskinsky H. "Cerenkov Radiation at Microwave Frequencies"
in advances in Electronics and Electron Physics, L. Marton,
(ed) Academic Press, New York, Vol. XIV, pp. 265, (1961).

7. Piestrup, M.A., R.A. Powell, G.B. Rothbart, C.K. Chen and R.H.
Fantell, Appl. Phys. Lett., 28. 92, (1976).

S. Vialsh, J.S., T.C. Marshall and 3. P. Schlesinger, Phvs . Fluids,
20 709 1977.

9. Valsh, J.E. "Cerenkov and Cerenkov-Raman Radiation Sources" in
Physics of Quantum Electronics, S. Jacobs, M. Sargent and

M. Scully (ed) Vol 7, Addison Wesley, reading, Mass., (1979).

10. Jackson, J.D., "Classical Electrodynamics", (2nc^ Edition)
John Wiley, New York 1975 (pp.4) Sec. 9.2 pp. 394ff.

11. Hofstadter , R. , Ann. Rev. Nuc Sci, Vol 7. pp. 2 31, (1937).

12. Casey, K.F., C. Yeh and Z.A. Keprielian, Phvs. Rev. 140 B 768
(1965) .

13. Hopf, F.A., P. Meystre and M.O. Scully, Phys. Rev. Lett 3_7
1215 (1976) .

22

r=TT/2Z

J!

*G

* e

Tf

^i

igure 1.

Qualitative Behavior of the Function I2(Z'). Both
the function G, from Eq. 25 in the text, and the
emission angle are displayed as independent
variables. G' and G" are upper and lower limits.

23

Pu

» OJ

Fiqure 2. Schematic Behavior of Power Emitted as a function of

Angular Frequency

24

F3 Eo

o -

— r

♦2 AY

W

o

i — i —

2TT

-5- ^;

Figure 3. Structure of charge pulse from a travelling wave
accelerator. y is the phase angle of an electron
relative to the peak of the travelling wave
accelerating field. Electrons in the range +A^are
passed by magnetic deflection system.

25

APPENDIX A
DERIVATION OF CERENKOV RADIATION FOR A SINGLE PULSE OF
CHARGE .

Let the pulse be described by

p' (r,t) = p 'Q(r- vt) Al

Both kz and go are continuous variables in this case; "v is again
along the z axis. If we expand in terras of a four dimensional
fourier integral,

p'(r,t) = l/(2ir)4 e1^ " k-r) o ' (k, w) d3kdoj A2

It may be shown that the condition Ai gives:

p'(k,aj) = 2tt6(o) - kzv) p'Q(k) A3

where p' (k) is the three dimensional spatial tranform of p1
evaluated at t = o. All the fields have fourier integral rather
than fouries series expansions and the energy radiated per unit
solid angle become

2 ^ ~ £ X ]- -, u f 2,

dt n • S = t" ,,,,2 — (o dw
2tt (4tt) c J _

'Jd3r' J df ei(ct'-n.?')o)/c. x ?(*,tl)|

A4
m |2

W ("> , n ) do)

Al

The integrand is a symmetric function of co so that

W

<-'*-^3*"\ffl)r*w

icu(t' -n-r' /c) ^ £,«»■, . , ,
e nxJ(r!,t')

A5

where

1 y ,* -, 2.,2

= — -, — co vn x v) M

, - 3 c

lOTT

M =

,3 , , Mcot'-n-r /c) ,,-"", ,, \
d r d-c e p (r , t )

A6

Now we may write p'(r',t') in a fourier integral representation

•(r\t')

(2ir)

J.

*3. , , , ,,.-*, M -i(o) 't'-k' -r ')

a k dco p ( .< , co ) e

A7

Inserting eq . A3 into eq. A7 and the result into eq A6, the
integral over d^ ' involves only exponentials and yields
(2rr)3 o3(yv' _ om/c) , so that eq A6 becomes

II d3k'dco'ei(a)"a)")t,63(k' - ncu/c:

5 (aj'-k. 'v)p' (k')

Now the integral over go' may be done; because of the 6 function,

to' is evaluated at kz'v.

M =

/dt^AV

i i

L(co-k v) t £.3.^, a. . .,,-*..

z (S (k ' - con/c) p ' (k ' )

-v o

A2

Now do the integrals over kx ' , ky' and kz', noting that
k ' appears in the exponential, but k ' and k ' do not,

u _ [,, icot' -iwt'n
M = / dt e e :

v/c ' / / / / s

; p 0 (am /c,am /C/um /c)
~ x y z

This may be written as

M = dt e p0 (oon/c) A8

where

H = i-n v/c

z A9

If we let the time interval be finite, from -T to + T, the
integral is easily done:

2_

iM = — sin ojHTPo (nco/c) A10

M = 4T sin cjHtI Po (noo/c) | A]_l

(coHT) 2

A3

This result, eq. All may be inserted into A5 for oo • The
factor n x v is just sin 9 where 9 is the angle between the
radiation and the beam axis.

W(oj,n) = —- — -H- (a sinZ9 4TZ sin coHT'l p ' (nca/c) | Al3

16 * C (wHT)2 '

W is the energy radiated per unit solid angle per unit
angular frequency, co • To proceed to the total energy, multiply
by d ft (solid angle) and integrate. But nz = cos 9 so that dQ
may be related to cH :

d ft = d(cos 8)dft= - — dHdft A , ,,

v A14

Tii'i functions in eq . A13 do not contain <j> so that integration
over G yields 2 it. Thus:

wi , jn 1 y 2 m2f . 2 , ',2 sin coHT ,„

A(co,nidft = — co T /sin 9|p 0 | dH 5

2rr2 ~ (ooHT)2

The sm2 goHT / (coHT) 2 factor in the integral is peaked at H = 0,

c
which by eq. A9 is at nz = cos 9 = — , or the usual Cerenkov

angle, 9 c. This function is more strongly peaked about H = 0

for large values of T, and in fact, for large T we may evaluate

sin^ and P '0 at the point corresponding to H =0. Then the

integral

oo

J

2 2

dx sin (ax) /(ax) = rr/a

A4

may be used to evaluate eq . A15, yielding
ffwdfi = i- -| 2Tsin29c|p;(Vc)|2

JJ

A16

The emmission was assumed to occur in a time interval from -T to
•KT ; accordingly dividing by 2T yields a rate of emmission, and
multiplying by v converts to emmission per unit path length. Thus
we obtain, for the large T limit:

dxdco

GO) =

— ujdco sin 8 |o'(nco/c)|

A17

where d^E/^^ is the energy emitted per unit path length
per unit angular frequency range co .

The corresponding expression for T not. large is

H"

d2E , U , /coT

-= — -=— dco = -: — go dco —
dxdco 4tt it

sin 9jp * (noj/c) J sin^coHT
H' (coHT)

Aia

where H" and H' are the value of H corresponding to 9=0 and
5 = tt respectively.

Equations Al 7 and A18 then describe the energy radiated per
unit path length and per unit angular frequency range. For the
non periodic (single) pulse the radiation has a continuous
freauency spectrum. For a point charae q, p' (k) is identically q
and the usual Cerenkov formula is obtained. Equation A17 is
quoted by Jelly, but only with the form factor corresponding to a
uniform line charge of length L' .

APPENDIX B
DERIVATION OF EQUATION 7

Equation 7 is derived for the case in which J(r,t) is

— »
expanded in fourier series. Let the fourier coefficient for A

be given by:

T

r

A(r\o)) = i J dt A(r\t) eicut B1

o

Assume that the green's function solution for A(r,t) is given:

(r,t) = ujr/Td3r' |

dt' J(r',t) D(r - r', t - t)

32

where

D(r,t) - ~ Sit - r/c)

Let the current density be expanded in a fourier series

B3

J(?\f ) =Z e-iw't,J(?',aJ) B4

co'

Then insert 32, B3 and B4 into 31 to obtain

t (*.«> = H. J dt e1^ JjfdV Jc.f ^ ,^— p-

B5

e ia),t' J(r« ,03')

_>

z

Bl

Do
-ioi 't'

Jl dt ' , note that t' appears in the <5 function and in
The result is t' is evaluated at t' = t - j r-r1 l/c.

T

u

A(r,co) = —

~ T

dt e

loot

dJr'

1 1

4tt i-» •»'

r-r

B6

I

0)

-ico 't ico ' j r-r ' I /c ^, -^, 1,
e e ' ' J(r ,a> )

Do the integral on t, note that

T

kj dt •"

L (co-co ') L

= 6

CO CO

B7

Then do the sum on co '

A(r ,co; =

4tt

d r

J(r ,co; e

ico I r-r 1 /c

r-r

38

This proves the desired result, B8 is equation 7 as used in
the main text.

Appendix C

TEMPORAL STRUCTURE OF THE ELECTRON PULSE FROM A TRAVELLING
WAVE ACCELERATOR.

Assume that the energy of a single electron emerging from a
linac with phase $ relative to the travelling wave field is

CI

E = E cos-J;
o

This relation is shown on fig. 3, along with some dots
representing electrons rear the maximum energy Ec, with phases
clustered about 'b = o and ty - 2^. Two bunches, separated by a
phase difference of 2t, are separated by a time T-i = 1/f where
f is the accelerator frequency, which is fQ = 2.85 x 10^ Hz
for a typical S-band accelerator of the Stanford type.

If a deflection system with energy resolution slit passes
only energies E from EQ to EQ- E the corresponding range of phase

Aip is

C2

AE = E - E = E (1- cosAijj)
o c

For AiJj small, this reduces to

AE = (Ai|;) 2 c3

Eo " 2

CI

The temporal pulse length T2 is

Tn = 2A<|; T,/2

C4

2 x

or

T_ = T • 2u^/2tt

If C3 is used to evaluate Aii> in terms of the fractional
energy resolution AE/EQ

O 7T CD

For 1% energy resolution, 1,2/T1 is about 1/20. The
electrons thus emerge in short bunches, and the charge and
current, when expressed in a fourier expansion, should have very
strong harmonic content up to and aoove the 20th harmonic.

C2.

Appendix D
FORM FACTORS

This section provides details and examples of form factors
for various charge distributions. From the main text, F differs
only from P, the fourier transform of P, by the total charge q of
the bunch, so that for k = o, F reduces to unity. Thus we define

£ ///d3r P(r)ei^'"

F(k) = £ / M dJr Pirje^1 L

q

For spherically symmetric charge distributions , let k*r = kru,
where u is the cosine of the angle between k and r. In spherical
coordinates, d^r = dudyr2dr. Then we find,

CO

F ( k ) =1±I / dr r p ( r )

r~< 1, -J

sin Kr D2

q k

For k very small, sin x may be replaced by x - x 3/6 and we
have

*°

F(k) = 1 ±1 / dr r p(r)[kr - k3r3/6] D3

q k J

Then the two terms in the square bracket lead to separate
integrals, the first term being unity and the second is similar to
the integral used to calculate the mean square radius, <r~>,
exceot for a factor k2/6. Thus we have

F(k) = 1 - k2<r2>/6

D\

D4

For a uniform spherical charge distribution of radius R, as
well a a spherical shell of radius k, the integral D2 may be
performed easily

F(k) = 3 (sin kR - kR cos kR)

(Solid sphere)

(kR)T

D5

F(k) = I sin(kR) (Spherical shell)
kR

D6

For a line charge concentrated on the z axis, we may return
to Dl and let p(r) = 5(x) <5(y) p"(z), so that

F(k) = 1 dz p"(z) eikz (line charge)

D7

Q

F(k) - j sin ( _) (uniform line charge

kZ 2

of length Z)

D8

Distorted spherical symmetry may be said to occur if the
scale transformation z' = pz serves to make p spherically
symmetric in tine prime system. Let Fg be the form factor
calculated by 0 2 in the prime frame. It is simple to show that

F(kx, ky, kz) = Fs(k^, k^, k^/p) D9

DZ

DISTRIBUTION LIST

Office of Naval Research
CAPT. M.A. Howard, Code 210
800 N. Quincy
Arlington, VA. 22217

Office of Naval Research
CDR James Offutt
1030 E. Green St .
Pasadena, CA. 91106

Library,

Naval Postgraduate School

Office of Research Administration

NPS, Code 012A

Monterey, CA. 93940

F.R. Buskirk & J.R. Neighbours 20

Naval Postgraduate School
Physics Dept . , Code 61
Monterey, CA. 93940

DUDLEY

KNOX LIBRARY- RESEARCH , REPORTS

5 6853 01057876 8

U21041